DB2 LUW pureScale ONLINE Maintenenace: Applying a GPFS efix and/or PTF

Content

The following instructions can be used to to install the GPFS efix and/or PTF while allowing the DB2 Cluster to remain accessible (ONLINE Updates)::

Order of Maintenance

Maintenance should be applied in the following order:

• Member hosts
• Secondary CF host
• Please note that while this maintenance is being performed, the cluster has only one available CF. If that CF should fail, the cluster will fail.
• Primary CF host
• Please note that performing this maintenance on the Primary CF host will result in a CF failover to the secondary CF
• Please note that while this maintenance is being performed, the cluster has only one available CF. If that CF should fail, the cluster will fail.

GPFS efix and/or PTF update

The following is a README included with a typical GPFS efix. It has been modified for a pureScale environment. This section will be referenced in later steps.

Installation Instructions:
This efix has been packaged with epkg and needs to be applied using emgr.
The following website gives information and detailed command syntax
for the emgr command:

http://publib.boulder.ibm.com/infocenter/aix/v6r1/index.jsp?topic=%2Fcom.ibm.aix.cmds%2Fdoc%2Faixcmds2%2Femgr.htm

1. If you need to be at a particular ptf level (see prereq) before installing this efix go to
http://www-947.ibm.com/support/entry/portal/Downloads/Software/Other_Software/General_Parallel_File_System
and obtain fixes for your cluster.

2. Stop applications utilizing your GPFS filesystems    There is no need to stop the applications.

3. Unmount GPFS filesystems and shutdown the GPFS daemon.  More than likely , there will be nothing mounted.              
  - /usr/sbin/umount -t mmfs
  - /usr/lpp/mmfs/bin/mmshutdown
 
4. Verify the kernel extension has been unloaded (this step is not needed
   if you plan to reboot your nodes after installing service):             
   - /usr/lpp/mmfs/bin/mmfsenv -u

5. Upgrade GPFS on node by using the installp command SMIT if you need to be at higher ptf level
   before installing this efix.

6. To install the efix <package>.epkg.Z file, make the file available
   locally or via a NFS mount and issue the following command on
   all GPFS nodes:

   /usr/sbin/emgr -e <directory>/<package>.epkg.Z
 
   This will perform various steps and will apply the efix supplied.
   If errors are encountered, it will display them and abort
   the application and restore any necessary files.

7. Start GPFS daemon
   /usr/lpp/mmfs/bin/mmstartup

Performing maintenance on a member host

Repeat the following steps for each member host in the cluster.

You can perform maintenance on or apply updates to a member host without affecting the availability of the databases in the DB2® pureScale® instance.

Before you begin

Any DB2 server processes and any other processes accessing the file systems managed by DB2 cluster services must be shut down prior to running this task. All instances must be stopped on the host before the host can enter maintenance mode. When the instance is stopped on the host, the host is no longer a viable recovery target for failed members.
Note: You can only put one host at a time into maintenance mode while performing ONLINE updates.

To perform this task, you must be the DB2 cluster services administrator.

Procedure

1. As an instance user, perform the following steps:

a. Perform a member drain operation using the QUIESCE parameter of the db2stop command, as follows:

db2stop member member-id quiesce 30

where member-id represents the member that you want to put into maintenance mode.
For more information on quiescing a member, see Quiescing a member in the information center.

b. Before you stop the host, make sure that there is only the member targeted for maintenance on the host. That is, there must be no "failed over" member also executing on this host. Stop the instance on the member host, as follows:

db2stop instance on host-name

where host-name represents the name of the host of the given member.

2. As the DB2 cluster services administrator (typically root), perform the following steps:

a. Put the cluster manager into maintenance mode on the host, by issuing the following command:

DB2DIR/bin/db2cluster -cm -enter -maintenance

where DB2DIR represents the installation location of your DB2 copy.

b. Put the cluster file system service on the host into maintenance mode, by issuing the following command:

DB2DIR/bin/db2cluster -cfs -enter -maintenance

where DB2DIR represents the installation location of your DB2 copy.

c. Perform maintenance and reboot as necessary.

Install gpfs efix and/or PTF as per above Section "GPFS efix and/or PTF update"

What to do next

1. As the DB2 cluster services administrator, perform the following steps:

a. Perform any maintenance activities planned in addition to applying the GPFS efix and/or PTF. For example, make hardware configuration changes, reboot server if necessary, etc.

b. Exit cluster manager maintenance mode on the host by issuing the following command:

DB2DIR/bin/db2cluster -cm -exit -maintenance

where DB2DIR represents the installation location of your DB2 copy.

c. Exit cluster file system maintenance mode on the host by issuing the following command:

DB2DIR/bin/db2cluster -cfs -exit -maintenance

where DB2DIR represents the installation location of your DB2 copy.

2. As an instance user, perform the following steps:

a. Restart the instance on the host by issuing the following command:

db2start instance on host-name

where host-name represents the name of the host of the given member or CF.

b. Restart the member by issuing the following command:

db2start member member-id

where member-id represents the member that you want to put into maintenance mode.

Performing maintenance on a CF host

Repeat the following steps for each CF host in the cluster.

You can perform maintenance on or apply updates to a CF host without affecting the availability of the databases in the DB2® pureScale® instance.

Before you begin

Any DB2 server processes and any other processes accessing the file systems managed by DB2 cluster services must be shut down prior to running this task. All instances must be stopped on the host before the host can enter maintenance mode. When the instance is stopped on the host, the host is no longer a viable recovery target for failed CFs.
Note: You can only put one host at a time into maintenance mode when performing ONLINE updates.

To perform this task, you must be the DB2 cluster services administrator.

Procedure

1. As an instance user, perform the following steps:

a. Stop the CF executing on this host as follows:

db2stop cf cf-id

where cf-id represents the CF that you want to put into maintenance mode.

b. Before you stop the host, make sure that there are only the CFs targeted for maintenance on the host. Stop the instance on the CF host, as follows:

db2stop instance on host-name

where host-name represents the name of the host of the given CF.

2. As the DB2 cluster services administrator (typically root), perform the following steps:

a. Put the cluster manager into maintenance mode on the host, by issuing the following command:

DB2DIR/bin/db2cluster -cm -enter -maintenance

where DB2DIR represents the installation location of your DB2 copy.

b. Put the cluster file system service on the host into maintenance mode, by issuing the following command:

DB2DIR/bin/db2cluster -cfs -enter -maintenance

where DB2DIR represents the installation location of your DB2 copy.

c. Perform maintenance and reboot as necessary.

Install gpfs efix and/or PTF as per above Section "GPFS efix and/or PTF".

 
What to do next

1. As the DB2 cluster services administrator, perform the following steps:
 

a. Perform any maintenance activities planned in addition to the GPFS maintenance. For example, make hardware configuration changes, reboot server if required, etc.

b. Exit cluster manager maintenance mode on the host by issuing the following command:

DB2DIR/bin/db2cluster -cm -exit -maintenance

where DB2DIR represents the installation location of your DB2 copy.

c. Exit cluster file system maintenance mode on the host by issuing the following command:

DB2DIR/bin/db2cluster -cfs -exit -maintenance

where DB2DIR represents the installation location of your DB2 copy.

2. As an instance user, perform the following steps:
 

a. Restart the instance on the host by issuing the following command:

db2start instance on host-name

where host-name represents the name of the host of the given CF.

b. Restart the CF by issuing the following command:

db2start cf cf-id

where cf-id represents the CF that you had placed in maintenance mode

c.This CF will start as the secondary CF. If there are active databases within the cluster , ensure this secondary CF attains PEER state.